New Delhi: Anthropic, the company behind the Claude chatbot, will start asking some users to prove their age or identity by submitting a government ID, a photo or video of themselves, and facial geometry data. The requirement is part of a revised privacy policy that takes effect 8 July.
The policy adds a new category called Verification Data to describe this process. Depending on which method of verification a user goes through, Anthropic says it may collect an image of a government-issued ID along with the details printed on it, such as the ID number and date of birth, plus a photo or video of the person and a facial geometry template generated from that image.
It also repeats that Anthropic’s services are not meant for anyone under 18 and asks anyone who learns that a child has submitted personal data to the company to email Anthropic so it can be reviewed and removed.
The company notes that facial geometry can count as biometric data under some laws, which generally means it receives extra legal protection.
Anthropic also keeps a record of the outcome e.g. whether the user meets the criteria to access a particular feature. This may, possibly, eliminate the need to submit these data repeatedly for future usage.
The second major change in its privacy policy addresses something that has quietly become central to how Claude works: its ability to connect to outside apps and carry out tasks on a user’s behalf.
After the latest changes, the policy distinguishes between “Inputs,” meaning anything a person submits to Claude, including files they upload and data pulled in from connected services, and “Outputs,” the responses and actions Claude generates in return.
Once a user links Claude to a third-party service through a connector, plugin, webhook, or API, Anthropic says Claude can send Inputs, Outputs, and instructions straight to that service to do things like read a file, send a message, or look something up. From that point, the outside service handles the data under its own privacy policy, and the connection can stay active until the user switches it off.
A new provision also covers people who take part in Anthropic’s research studies, surveys, or interviews. The company says it will collect their responses and combine them with their accounts and other details already shared with the platform for user-specific analysis.
The policy also says more about how Anthropic talks to its users, including that it may use account and usage data to send service updates and what it calls tailored recommendations about its products.
A new table lays out, for each purpose Anthropic processes data, which categories of data are involved and which legal basis it relies on, from contract terms to consent to legitimate interest.
On the points users tend to ask about most, Anthropic says nothing has changed: it does not sell personal data, Claude carries no ads, and users can still decide in their account settings whether their conversations help train its models.
There is one carve-out to that last point. Even if a user opts out, a conversation flagged for a safety review or sent in through a feedback tool can still be used to improve Anthropic’s models and its systems for catching harmful content.
The new rules apply only to Claude Free, Pro, and Max, the plans aimed at individual consumers. Claude Team, Claude Enterprise, the Claude Platform, and anything covered by a separate commercial agreement continue to run under their existing contracts.
The policy keeps separate sections for users in Canada, Brazil, and South Korea covering local consent rules, cross-border data transfers, and country-specific rights, and it names Anthropic Korea, Limited as its representative in Seoul.
It also repeats that Anthropic’s services are not meant for anyone under 18 and asks anyone who learns that a child has submitted personal data to the company to email Anthropic so it can be reviewed and removed.
The fight in US
Separately, and unrelated to this policy update, Anthropic has spent much of 2026 in a standoff with the US government over national security. In February, President Donald Trump told federal agencies to stop using Anthropic’s technology, and the Department of Defense, which now also goes by Department of War, labelled the company a supply chain risk.
Anthropic sued to fight the label, and in March a federal judge in the Northern District of California put the directive on hold with a preliminary injunction, after which the General Services Administration put Anthropic’s products back on its list of approved federal vendors.
The fight flared up again this month: Anthropic said US agencies ordered it to cut off foreign nationals from its newest models, Fable 5 and Mythos 5, citing a cybersecurity vulnerability and national security concerns.
The company complied but said it disagrees with the standard being applied, arguing it would effectively halt new frontier model releases across the industry if other companies were held to the same bar.
(Edited by Ajeet Tiwari)
Also Read: ‘AI will soon become so capable that I worry’: Anthropic CEO calls for urgent binding AI regulations